update_path

README.md

@@ -1,70 +1,146 @@
 ```markdown
-# finish.sh
+# finish
 
-AI-powered shell completion that runs 100 % on your machine.
+AI-powered shell completion that runs 100% on your machine.
 
 One command and your terminal learns what you type next.
 
 ## Install
 
 ```bash
+# On Ubuntu/Debian, first ensure python3-venv is installed
+sudo apt-get install python3-venv
+
+# Then run the installer
 curl -sSL https://git.appmodel.nl/tour/finish/raw/branch/main/docs/install.sh | bash
 source ~/.bashrc   # or ~/.zshrc
+
+# for plato special, add plato.lan to /etc/hosts:
+sudo nano /etc/hosts  -   192.168.1.74    plato.lan
 ```
 
-Press `Tab` twice on any partial command and finish.sh suggests the rest—no cloud, no data leak, no latency.
+Press **Alt+\\** after typing any command to get intelligent completions—no cloud, no data leak, instant results.
 
 ## How it works
 
-1. Captures your current directory, recent history, env vars.
-2. Builds a concise prompt for a local LLM (LM-Studio, Ollama, or any OpenAI-compatible endpoint).
-3. Returns ranked completions in <200 ms, cached for instant replay.
+1. Captures your current directory, recent history, env vars, and available tools
+2. Analyzes your intent and builds a context-aware prompt
+3. Queries your local LLM (LM Studio, Ollama, or any OpenAI-compatible endpoint)
+4. Returns 2-5 ranked completions with an interactive picker
+5. Results are cached for instant replay
 
-## Use
+## Usage
+
+Type a command, then press **Alt+\\**:
 
 ```bash
-docker <Tab><Tab>          # → docker run -it --rm ubuntu bash
-git commit <Tab><Tab>      # → git commit -m "feat: add finish.sh"
-# large files <Tab><Tab>   # → find . -type f -size +100M
+# Natural language commands
+show gpu status                    # → nvidia-smi
+resolve title of website google.com # → curl -s https://google.com | grep -oP '<title>\K[^<]+'
+make file about dogs               # → echo "About Dogs" > dogs.md
+
+# Partial commands
+git commit                         # → git commit -m "..."
+docker run                         # → docker run -it --rm ubuntu bash
+find large files                   # → find . -type f -size +100M
 ```
 
-Dry-run mode:
-
-```bash
-finish --dry-run "tar czf backup.tar.gz"
-```
+Navigate with ↑↓, press Enter to accept, Esc to cancel.
 
 ## Configure
 
+View current configuration:
+
 ```bash
-finish config set endpoint http://plato.lan:11434/v1/chat/completions
-finish config set model codellama:13b
-finish model               # interactive picker
+finish config
 ```
 
-## Providers
-
-| Provider  | Auth | URL                         | Notes   |
-|-----------|------|-----------------------------|---------|
-| LM-Studio | none | `http://localhost:1234/v1`  | default |
-| Ollama    | none | `http://localhost:11434`    |         |
-| OpenAI    | key  | `https://api.openai.com/v1` |         |
-
-Add others by editing `~/.finish/config`.
-
-## Commands
+Edit configuration file:
 
 ```bash
-finish install    # hook into shell
-finish remove     # uninstall
-finish clear      # wipe cache & logs
-finish usage      # tokens & cost
+nano ~/.finish/finish.json
+```
+
+## Configuration Examples
+
+### Local Ollama
+```json
+{
+  "provider": "ollama",
+  "model": "llama3:latest",
+  "endpoint": "http://localhost:11434/api/chat",
+  "temperature": 0.0,
+  "api_prompt_cost": 0.0,
+  "api_completion_cost": 0.0,
+  "max_history_commands": 20,
+  "max_recent_files": 20,
+  "cache_size": 100
+}
+```
+
+### LM Studio
+```json
+{
+  "provider": "lmstudio",
+  "model": "dolphin3.0-llama3.1-8b@q4_k_m",
+  "endpoint": "http://localhost:1234/v1/chat/completions",
+  "temperature": 0.0,
+  "api_prompt_cost": 0.0,
+  "api_completion_cost": 0.0,
+  "max_history_commands": 20,
+  "max_recent_files": 20,
+  "cache_size": 100
+}
+```
+
+### OpenAI (or compatible API)
+```json
+{
+  "provider": "lmstudio",
+  "model": "gpt-4",
+  "endpoint": "https://api.openai.com/v1/chat/completions",
+  "api_key": "sk-...",
+  "temperature": 0.0,
+  "api_prompt_cost": 0.03,
+  "api_completion_cost": 0.06,
+  "max_history_commands": 20,
+  "max_recent_files": 20,
+  "cache_size": 100
+}
 ```
 
 ## Requirements
 
-Bash ≥4 or Zsh ≥5, curl, jq, bc.  
-Optional: bash-completion.
+- Python 3.7+
+- Bash ≥4 or Zsh ≥5
+- A local LLM running (Ollama, LM Studio, etc.) or API access
+
+Python dependencies (installed automatically):
+- httpx
+- prompt_toolkit
+- rich
+
+## Commands
+
+```bash
+finish install    # Set up Alt+\ keybinding
+finish config     # Show current configuration
+finish command "text"  # Test completions manually
+```
+
+## Advanced
+
+### Debug mode
+```bash
+export FINISH_DEBUG=1
+finish command "your command here"
+cat ~/.finish/finish.log
+```
+
+### Clear cache
+```bash
+rm -rf ~/.finish/cache/*.json
+```
 
 ## License
 

docs/install.sh

@@ -4,7 +4,7 @@
 
 set -e
 
-ACSH_VERSION="v0.5.1"
+ACSH_VERSION="v0.6.0"
 BRANCH_OR_VERSION=${1:-main}
 REPO_URL="https://git.appmodel.nl/tour/finish/raw/branch"
 
@@ -97,11 +97,9 @@ main() {
     SHELL_TYPE=$(detect_shell)
     case "$SHELL_TYPE" in
         zsh)
-            SCRIPT_NAME="finish.zsh"
             RC_FILE="$HOME/.zshrc"
             ;;
         bash)
-            SCRIPT_NAME="finish.sh"
             RC_FILE="$HOME/.bashrc"
             ;;
         *)
@@ -112,6 +110,29 @@ main() {
 
     echo "Detected shell: $SHELL_TYPE"
 
+    # Check Python and venv
+    echo "Checking Python installation..."
+    if ! command -v python3 > /dev/null 2>&1; then
+        echo_error "Python 3 is required but not found"
+        echo "Install it with:"
+        echo "  Ubuntu/Debian: sudo apt-get install python3 python3-pip python3-venv"
+        echo "  macOS: brew install python3"
+        exit 1
+    fi
+    echo_green "✓ Python 3 found: $(python3 --version)"
+
+    # Check if venv module is available
+    if ! python3 -m venv --help > /dev/null 2>&1; then
+        echo_error "python3-venv module is not installed"
+        echo "Install it with:"
+        echo "  Ubuntu/Debian: sudo apt-get install python3-venv"
+        echo "  CentOS/RHEL: sudo yum install python3-venv"
+        echo "  macOS: (included with python3)"
+        exit 1
+    fi
+    echo_green "✓ python3-venv available"
+    echo ""
+
     # Check dependencies
     echo "Checking dependencies..."
     if ! check_dependencies; then
@@ -144,17 +165,73 @@ main() {
         exit 1
     fi
 
-    # Create directory if needed
+    # Create directories if needed
     mkdir -p "$(dirname "$INSTALL_LOCATION")"
+    mkdir -p "$HOME/.venvs"
 
-    # Download script
-    echo "Downloading finish.sh..."
-    URL="$REPO_URL/$BRANCH_OR_VERSION/$SCRIPT_NAME"
+    # Download Python script
+    echo "Downloading finish.py..."
+    URL="$REPO_URL/$BRANCH_OR_VERSION/src/finish.py"
     if ! download_file "$URL" "$INSTALL_LOCATION"; then
         echo_error "Failed to download from $URL"
         exit 1
     fi
 
+    # Download requirements.txt
+    echo "Downloading requirements.txt..."
+    TEMP_REQ="/tmp/finish_requirements.txt"
+    REQ_URL="$REPO_URL/$BRANCH_OR_VERSION/requirements.txt"
+    if ! download_file "$REQ_URL" "$TEMP_REQ"; then
+        echo_error "Failed to download requirements.txt from $REQ_URL"
+        exit 1
+    fi
+
+    # Create virtualenv and install dependencies
+    echo "Creating virtual environment..."
+    VENV_PATH="$HOME/.venvs/finish"
+    if [ -d "$VENV_PATH" ]; then
+        echo "Virtual environment already exists, removing old one..."
+        rm -rf "$VENV_PATH"
+    fi
+
+    if ! python3 -m venv "$VENV_PATH"; then
+        echo_error "Failed to create virtual environment"
+        echo "Make sure python3-venv is installed:"
+        echo "  Ubuntu/Debian: sudo apt-get install python3-venv"
+        exit 1
+    fi
+    echo_green "✓ Virtual environment created at $VENV_PATH"
+
+    echo "Installing Python dependencies..."
+    if ! "$VENV_PATH/bin/pip" install --quiet --upgrade pip; then
+        echo_error "Failed to upgrade pip"
+        exit 1
+    fi
+
+    if ! "$VENV_PATH/bin/pip" install --quiet -r "$TEMP_REQ"; then
+        echo_error "Failed to install dependencies"
+        cat "$TEMP_REQ"
+        exit 1
+    fi
+    echo_green "✓ Dependencies installed"
+    rm -f "$TEMP_REQ"
+
+    # Update shebang to use venv python (compatible with macOS and Linux)
+    if sed --version >/dev/null 2>&1; then
+        # GNU sed (Linux)
+        sed -i "1s|.*|#!$VENV_PATH/bin/python3|" "$INSTALL_LOCATION"
+    else
+        # BSD sed (macOS)
+        sed -i '' "1s|.*|#!$VENV_PATH/bin/python3|" "$INSTALL_LOCATION"
+    fi
+
+    # Verify shebang was updated correctly
+    SHEBANG=$(head -n 1 "$INSTALL_LOCATION")
+    if [ "$SHEBANG" != "#!$VENV_PATH/bin/python3" ]; then
+        echo_error "Failed to update shebang. Got: $SHEBANG"
+        exit 1
+    fi
+
     chmod +x "$INSTALL_LOCATION"
     echo_green "✓ Installed to $INSTALL_LOCATION"
     echo ""
@@ -183,8 +260,18 @@ main() {
         fi
     fi
 
-    # Run finish install
-    echo "Running finish installation..."
+    # Test that finish works
+    echo "Testing installation..."
+    if ! "$INSTALL_LOCATION" --version > /dev/null 2>&1; then
+        echo_error "finish command failed to execute"
+        echo "Testing with direct python call..."
+        "$VENV_PATH/bin/python3" "$INSTALL_LOCATION" --version
+        exit 1
+    fi
+    echo_green "✓ finish command works"
+
+    # Run finish install to set up keybinding
+    echo "Setting up shell keybinding..."
     if "$INSTALL_LOCATION" install; then
         echo ""
         echo_green "=========================================="
@@ -195,10 +282,18 @@ main() {
         echo "  1. Reload your shell configuration:"
         echo "     source $RC_FILE"
         echo ""
-        echo "  2. Select a language model:"
-        echo "     finish model"
+        echo "  2. Configure your LLM endpoint:"
+        echo "     finish config"
         echo ""
-        echo "  3. Start using by pressing Tab twice after any command"
+        echo "  3. Edit config if needed:"
+        echo "     nano ~/.finish/finish.json"
+        echo ""
+        echo "  4. Start using by pressing Alt+\\ after typing a command"
+        echo ""
+        echo "Example:"
+        echo "  Type: show gpu status"
+        echo "  Press: Alt+\\"
+        echo "  Select completion with ↑↓ arrows, Enter to accept"
         echo ""
         echo "Documentation: https://git.appmodel.nl/tour/finish"
     else

requirements.txt

@@ -1,4 +1,3 @@
 httpx
 prompt_toolkit
 rich
-hashlib

src/finish.py

@@ -1,4 +1,4 @@
-#!/home/mike/.venvs/cli/bin/python3
+#!/usr/bin/env python3
 """
 finish.py  –  AI shell completions that never leave your machine.
 """
@@ -73,37 +73,91 @@ def _recent_files() -> str:
     except Exception:
         return ""
 
-def _help_for(cmd: str) -> str:
-    try:
-        return subprocess.check_output([cmd, "--help"], stderr=subprocess.STDOUT, timeout=2).decode()
-    except Exception:
-        return f"{cmd} --help not available"
+def _installed_tools() -> str:
+    """Check for commonly used tools that might be relevant"""
+    tools = ["curl", "wget", "jq", "grep", "awk", "sed", "python3", "node", "docker", "git", "nvcc", "nvidia-smi"]
+    available = []
+    for tool in tools:
+        try:
+            subprocess.run(["which", tool], capture_output=True, timeout=0.5, check=True)
+            available.append(tool)
+        except Exception:
+            pass
+    return ", ".join(available) if available else "standard shell tools"
+
+def _get_context_info() -> dict:
+    """Gather comprehensive shell context"""
+    return {
+        "user": os.getenv("USER", "unknown"),
+        "pwd": os.getcwd(),
+        "home": os.getenv("HOME", ""),
+        "hostname": os.getenv("HOSTNAME", os.getenv("HOST", "localhost")),
+        "shell": os.getenv("SHELL", "bash"),
+        "tools": _installed_tools(),
+    }
 
 def build_prompt(user_input: str) -> str:
+    """Build an enhanced prompt with better context and instructions"""
     c = cfg()
-    term_info = f"""User: {os.getenv("USER")}
-PWD:  {os.getcwd()}
-HOME: {os.getenv("HOME")}
-HOST: {os.getenv("HOSTNAME")}
-SHELL: bash"""
-    prompt = f"""You are a helpful bash-completion script.
-Generate 2–5 concise, valid bash commands that complete the user’s intent.
+    ctx = _get_context_info()
 
-Reply **only** JSON: {{"completions":["cmd1","cmd2",...]}}
+    # Analyze user intent
+    user_words = user_input.lower().split()
+    intent_hints = []
+    if any(word in user_words for word in ["resolve", "get", "fetch", "download", "curl", "wget"]):
+        intent_hints.append("The user wants to fetch/download data")
+    if any(word in user_words for word in ["website", "url", "http", "html", "title"]):
+        intent_hints.append("The user is working with web content")
+    if any(word in user_words for word in ["parse", "extract", "grep", "find"]):
+        intent_hints.append("The user wants to extract/parse data")
+    if any(word in user_words for word in ["create", "make", "new", "write", "edit"]):
+        intent_hints.append("The user wants to create or edit files")
+    if any(word in user_words for word in ["file", "document", "text", "story", "about"]):
+        intent_hints.append("The user is working with files/documents")
+    if any(word in user_words for word in ["install", "setup", "configure", "apt", "pip"]):
+        intent_hints.append("The user wants to install or configure software")
+    if any(word in user_words for word in ["gpu", "nvidia", "cuda", "memory", "cpu"]):
+        intent_hints.append("The user wants system/hardware information")
 
-User command: {user_input}
+    intent_context = "\n".join(f"- {hint}" for hint in intent_hints) if intent_hints else "- General command completion"
 
-Terminal context:
-{term_info}
+    prompt = f"""You are an intelligent bash completion assistant. Analyze the user's intent and provide practical, executable commands.
 
-History:
+USER INPUT: {user_input}
+
+DETECTED INTENT:
+{intent_context}
+
+SHELL CONTEXT:
+- User: {ctx['user']}
+- Working directory: {ctx['pwd']}
+- Hostname: {ctx['hostname']}
+- Available tools: {ctx['tools']}
+
+RECENT COMMAND HISTORY:
 {_sanitise_history()}
 
-Recent files:
-{_recent_files()}
+INSTRUCTIONS:
+1. Understand what the user wants to accomplish (not just the literal words)
+2. Generate 2-5 practical bash commands that achieve the user's goal
+3. Prefer common tools (curl, wget, grep, awk, sed, jq, python)
+4. For file creation: use echo with heredoc, cat, or nano/vim
+5. For web tasks: use curl with headers, pipe to grep/sed/awk for parsing
+6. For complex parsing: suggest one-liners with proper escaping
+7. Commands should be copy-paste ready and executable
+8. Order by most likely to least likely intent
+
+EXAMPLES:
+- "resolve title of website example.com" → curl -s https://example.com | grep -oP '<title>\\K[^<]+'
+- "show gpu status" → nvidia-smi
+- "download json from api.com" → curl -s https://api.com/data | jq .
+- "make a new file story.txt" → nano story.txt
+- "create file and write hello" → echo "hello" > file.txt
+- "write about dogs to file" → cat > dogs.txt << 'EOF' (then Ctrl+D to finish)
+
+OUTPUT FORMAT (JSON only, no other text):
+{{"completions":["command1", "command2", "command3"]}}"""
 
-Help:
-{_help_for(user_input.split()[0])}"""
     return prompt
 
 # --------------------------------------------------------------------------- #
@@ -155,9 +209,9 @@ async def llm_complete(prompt: str) -> List[str]:
             f.write(f"\n=== {time.strftime('%Y-%m-%d %H:%M:%S')} ===\n")
             f.write(f"Raw response:\n{raw}\n")
 
-    # try json first
+    # try json first - use strict=False to be more lenient
     try:
-        result = json.loads(raw)["completions"]
+        result = json.loads(raw, strict=False)["completions"]
         if os.getenv("FINISH_DEBUG"):
             with open(LOG_FILE, "a") as f:
                 f.write(f"Parsed completions: {result}\n")
@@ -166,8 +220,28 @@ async def llm_complete(prompt: str) -> List[str]:
         if os.getenv("FINISH_DEBUG"):
             with open(LOG_FILE, "a") as f:
                 f.write(f"JSON parse failed: {e}\n")
+                f.write(f"Attempting manual extraction...\n")
+
+        # Try to extract JSON array manually using regex
+        try:
+            match = re.search(r'"completions"\s*:\s*\[(.*?)\]', raw, re.DOTALL)
+            if match:
+                # Extract commands from the array - handle escaped quotes
+                array_content = match.group(1)
+                # Find all quoted strings
+                commands = re.findall(r'"([^"\\]*(?:\\.[^"\\]*)*)"', array_content)
+                if commands:
+                    if os.getenv("FINISH_DEBUG"):
+                        with open(LOG_FILE, "a") as f:
+                            f.write(f"Manual extraction succeeded: {commands}\n")
+                    return commands[:5]
+        except Exception as e2:
+            if os.getenv("FINISH_DEBUG"):
+                with open(LOG_FILE, "a") as f:
+                    f.write(f"Manual extraction failed: {e2}\n")
+
         # fallback: grep command-like lines
-        fallback = [ln for ln in raw.splitlines() if re.match(r"^(ls|cd|find|cat|grep|echo|mkdir|rm|cp|mv|pwd|chmod|chown)\b", ln)][:5]
+        fallback = [ln for ln in raw.splitlines() if re.match(r"^(ls|cd|find|cat|grep|echo|mkdir|rm|cp|mv|pwd|chmod|chown|nano|vim|touch)\b", ln)][:5]
         if os.getenv("FINISH_DEBUG"):
             with open(LOG_FILE, "a") as f:
                 f.write(f"Fallback completions: {fallback}\n")